More tests for PyAction in parallel

[lisajulia]

Currently failing now:

• mpi.compareSeparateECLFiles_flow+pyaction_gconprod_insert_kw_4_procs
• mpi.compareSeparateECLFiles_flow+pyaction_gruptree_insert_kw_4_procs
• mpi.compareSeparateECLFiles_flow+pyaction_wconprod_insert_kw_4_procs
• mpi.compareSeparateECLFiles_flow+pyaction_wefac_insert_kw_4_procs (this one is different)

Probably the first three are all the same error, in Pyaction I get this output:

Load balancing distributes 2794 active cells on 4 processes as follows:

rank owned cells overlap cells total cells

 0           768             198           966  
 1           724             125           849  
 2           667             165           832  
 3           635             116           751  

---

sum          2794             604          3398

Load balancing distributed the wells as follows:

well name ranks with perforated cells

        OP01       1  
        OP02       3  
        OP03       2

Well GI01 is inactive, with perforations on ranks: 0 1 2  
Well WI01 is inactive, with perforations on ranks: 0

Whereas the ACTIONX version gives me:

Load balancing distributes 2794 active cells on 4 processes as follows:

rank owned cells overlap cells total cells

 0           707             248           955  
 1           765             172           937  
 2           744             103           847  
 3           578             177           755  

---

sum          2794             700          3494

Load balancing distributed the wells as follows:

well name ranks with perforated cells

        GI01       2  
        OP01       1  
        OP02       0  
        OP03       3  
        WI01       0

so the error might have sth to do with the wells GI01 and WI01

[lisajulia]

jenkins build this please

[blattms]

Thanks a lot. I have created a PR replacing this one.

That PR works around the problem by allowing all wells to be distributed. The problem was that with PYACTION some wells appear to be inactive and are allowed to be distributed, anyway. This leads to different partitionings and the problem.